classifyre-cli 0.4.58__tar.gz → 0.4.60__tar.gz

classifyre_cli-0.4.60/.turbo/turbo-build.log

@@ -0,0 +1,3 @@
+$ uv sync
+Resolved 300 packages in 341ms
+Checked 51 packages in 1ms

{classifyre_cli-0.4.58 → classifyre_cli-0.4.60}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: classifyre-cli
-Version: 0.4.58
+Version: 0.4.60
 Summary: Classifyre CLI — scan and classify unstructured data sources
 License: MIT
 Keywords: data,ingestion,metadata,pii,secrets,unstructured

{classifyre_cli-0.4.58 → classifyre_cli-0.4.60}/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@classifyre/cli",
-  "version": "0.4.58",
+  "version": "0.4.60",
   "private": true,
   "scripts": {
     "build": "uv sync",

{classifyre_cli-0.4.58 → classifyre_cli-0.4.60}/pyproject.toml

@@ -1,6 +1,6 @@
 [project]
 name = "classifyre-cli"
-version = "0.4.58"
+version = "0.4.60"
 description = "Classifyre CLI — scan and classify unstructured data sources"
 readme = "README.md"
 requires-python = ">=3.12"

{classifyre_cli-0.4.58 → classifyre_cli-0.4.60}/src/pipeline/worker_pool.py

@@ -45,6 +45,13 @@ def _init_worker_threads(threads_per_worker: int) -> None:
     threads = max(1, threads_per_worker)
     for var in ("OMP_NUM_THREADS", "MKL_NUM_THREADS", "OPENBLAS_NUM_THREADS"):
         os.environ.setdefault(var, str(threads))
+    # Detector work is background work: on a shared host (desktop) the pool
+    # must yield to the UI, so deprioritize workers where the OS allows it.
+    if hasattr(os, "nice"):
+        try:
+            os.nice(5)
+        except OSError:
+            pass
     try:
         import torch
 
@@ -227,7 +234,15 @@ class DetectorWorkerPool:
             mp_start_method = "forkserver" if "forkserver" in available else "spawn"
 
         ctx = multiprocessing.get_context(mp_start_method)
-        threads_per_worker = max(1, get_effective_cpu_count() // effective_workers)
+        # CLASSIFYRE_WORKER_THREADS caps BLAS/torch threads per worker; the
+        # desktop app sets it so workers*threads stays below the core count
+        # instead of saturating the machine. Unset (K8s/dev) keeps the
+        # divide-the-cores behavior.
+        threads_env = os.environ.get("CLASSIFYRE_WORKER_THREADS", "")
+        if threads_env.isdigit() and int(threads_env) > 0:
+            threads_per_worker = min(int(threads_env), 16)
+        else:
+            threads_per_worker = max(1, get_effective_cpu_count() // effective_workers)
         self._pool = ProcessPoolExecutor(
             max_workers=effective_workers,
             mp_context=ctx,

{classifyre_cli-0.4.58 → classifyre_cli-0.4.60}/src/sources/google_workspace/source.py

@@ -178,9 +178,7 @@ class GoogleWorkspaceSource(BaseSource):
 
         if isinstance(required, GoogleWorkspaceRequiredServiceAccount):
             if not isinstance(masked, GoogleWorkspaceMaskedServiceAccount):
-                raise ValueError(
-                    "service_account auth requires masked.service_account_json"
-                )
+                raise ValueError("service_account auth requires masked.service_account_json")
         elif isinstance(required, GoogleWorkspaceRequiredOAuth):
             if not isinstance(masked, GoogleWorkspaceMaskedOAuth):
                 raise ValueError(
@@ -655,9 +653,7 @@ class GoogleWorkspaceSource(BaseSource):
                         mime_type = export_mime
                         exported_as = export_mime
                 elif ref.size and ref.size > self._max_object_bytes():
-                    parse_error = (
-                        f"File size {ref.size} exceeds limit {self._max_object_bytes()}"
-                    )
+                    parse_error = f"File size {ref.size} exceeds limit {self._max_object_bytes()}"
                 else:
                     service = self._get_service()
                     request = service.files().get_media(fileId=ref.file_id)
@@ -792,9 +788,7 @@ class GoogleWorkspaceSource(BaseSource):
                         drive_id=_MY_DRIVE_ID, display_name="My Drive", drive_type="my_drive"
                     )
                 else:
-                    drive_ref = known_drives.get(drive_id) or self._resolve_shared_drive(
-                        drive_id
-                    )
+                    drive_ref = known_drives.get(drive_id) or self._resolve_shared_drive(drive_id)
                 ensure_drive(drive_ref)
                 if len(batch) >= self.BATCH_SIZE:
                     yield batch
@@ -849,7 +843,9 @@ class GoogleWorkspaceSource(BaseSource):
 
             logger.info(
                 "Processing %d/%d items from drive '%s'",
-                len(sampled), len(files), drive_id,
+                len(sampled),
+                len(files),
+                drive_id,
             )
 
             for file_ref in sampled:
@@ -884,18 +880,14 @@ class GoogleWorkspaceSource(BaseSource):
                 and bool(required.delegated_subject)
             )
             if uses_identity:
-                data = service.about().get(fields="user").execute(
-                    num_retries=self._max_retries()
-                )
+                data = service.about().get(fields="user").execute(num_retries=self._max_retries())
                 user = data.get("user", {})
                 result["status"] = "SUCCESS"
                 result["message"] = (
                     f"Connected to Google Drive API as {user.get('emailAddress', 'unknown')}"
                 )
             else:
-                data = service.drives().list(pageSize=1).execute(
-                    num_retries=self._max_retries()
-                )
+                data = service.drives().list(pageSize=1).execute(num_retries=self._max_retries())
                 drives = data.get("drives", [])
                 result["status"] = "SUCCESS"
                 if drives:
@@ -904,7 +896,9 @@ class GoogleWorkspaceSource(BaseSource):
                         f"{drives[0].get('name', 'unknown')}"
                     )
                 else:
-                    result["message"] = "Connected to Google Drive API. No shared drives accessible."
+                    result["message"] = (
+                        "Connected to Google Drive API. No shared drives accessible."
+                    )
         except Exception as exc:
             result["status"] = "FAILURE"
             result["message"] = f"Failed to connect to Google Drive API: {exc}"

{classifyre_cli-0.4.58 → classifyre_cli-0.4.60}/src/sources/youtube/source.py

@@ -339,8 +339,7 @@ class YouTubeSource(BaseSource):
         # Requiring both codecs avoids yt-dlp's ffmpeg merge path. Visual OCR only
         # needs readable text, while the audio track remains suitable for Whisper.
         opts["format"] = (
-            "best[height<=480][acodec!=none][vcodec!=none]/"
-            "best[acodec!=none][vcodec!=none]"
+            "best[height<=480][acodec!=none][vcodec!=none]/best[acodec!=none][vcodec!=none]"
         )
         opts["noplaylist"] = True
         opts["outtmpl"] = str(dest_dir / "%(id)s.%(ext)s")

{classifyre_cli-0.4.58 → classifyre_cli-0.4.60}/src/utils/legacy_office.py

@@ -106,9 +106,17 @@ def convert_legacy_office(
             # profile's lock file, which otherwise makes concurrent (or
             # crashed) soffice runs fail with "another instance is running".
             profile_dir = temp_path / "profile"
+            # --headless alone still lets macOS surface the app (Dock icon /
+            # brief window flash) on some LibreOffice builds; the extra flags
+            # suppress the start center, first-run wizard, and lock checks so
+            # conversions stay invisible.
             command = [
                 soffice,
                 "--headless",
+                "--invisible",
+                "--nologo",
+                "--nodefault",
+                "--nolockcheck",
                 "--norestore",
                 f"-env:UserInstallation={profile_dir.as_uri()}",
                 "--convert-to",

{classifyre_cli-0.4.58 → classifyre_cli-0.4.60}/src/utils/resources.py

@@ -62,4 +62,15 @@ def get_effective_memory_mb() -> int:
     except (FileNotFoundError, OSError, ValueError):
         pass
 
+    # macOS (and other POSIX without /proc): total physical memory via sysconf.
+    # Without this the desktop app fell through to the 4096 MB guess, which
+    # oversized the pool on small laptops and undersized it on big ones.
+    try:
+        pages = os.sysconf("SC_PHYS_PAGES")
+        page_size = os.sysconf("SC_PAGE_SIZE")
+        if pages > 0 and page_size > 0:
+            return max(256, (pages * page_size) // (1024 * 1024))
+    except (AttributeError, ValueError, OSError):
+        pass
+
     return 4096

{classifyre_cli-0.4.58 → classifyre_cli-0.4.60}/tests/test_google_workspace_source.py

@@ -144,9 +144,7 @@ class TestSampling:
         import json
         import os
 
-        encoded = base64.b64encode(
-            json.dumps(cursor_after_first).encode("utf-8")
-        ).decode("utf-8")
+        encoded = base64.b64encode(json.dumps(cursor_after_first).encode("utf-8")).decode("utf-8")
         os.environ["CLASSIFYRE_SAMPLING_CURSOR"] = encoded
         try:
             source2 = GoogleWorkspaceSource(recipe)
@@ -366,9 +364,7 @@ class TestEndToEndExtraction:
 
         # drives().list() -> one shared drive
         drives_list_result = {"drives": [{"id": "drive-1", "name": "Team Drive"}]}
-        mock_service.drives.return_value.list.return_value.execute.return_value = (
-            drives_list_result
-        )
+        mock_service.drives.return_value.list.return_value.execute.return_value = drives_list_result
 
         # files().list() -> root has one subfolder + nothing else, subfolder has
         # one binary file and one Google Doc.

{classifyre_cli-0.4.58 → classifyre_cli-0.4.60}/tests/test_kafka_source.py

@@ -15,6 +15,7 @@ CURSOR_ENV = "CLASSIFYRE_SAMPLING_CURSOR"
 def _encode_cursor(cursor: dict[str, Any]) -> str:
     return base64.b64encode(json.dumps(cursor).encode()).decode()
 
+
 # ── confluent-kafka fakes ────────────────────────────────────────────────
 
 

{classifyre_cli-0.4.58 → classifyre_cli-0.4.60}/tests/utils/test_video_processing.py

@@ -63,9 +63,7 @@ def test_changed_frame_ocr_is_timestamped_and_deduplicated(
     ocr_results = iter(["Opening slide", "Architecture slide"])
     monkeypatch.setattr(video_processing, "_ocr_text", lambda _frame: next(ocr_results))
 
-    segments = list(
-        video_processing.iter_video_ocr_segments(b"video", file_name="lecture.mp4")
-    )
+    segments = list(video_processing.iter_video_ocr_segments(b"video", file_name="lecture.mp4"))
 
     assert segments == [
         "[On-screen text 00:00:00]\nOpening slide",
@@ -122,9 +120,7 @@ def test_rapidocr_filters_low_confidence_lines(monkeypatch: pytest.MonkeyPatch)
     assert video_processing._ocr_text(object()) == "clear text\nanother line"
 
 
-def test_path_ocr_reuses_existing_video(
-    monkeypatch: pytest.MonkeyPatch, tmp_path: Path
-) -> None:
+def test_path_ocr_reuses_existing_video(monkeypatch: pytest.MonkeyPatch, tmp_path: Path) -> None:
     video_path = tmp_path / "video.mp4"
     video_path.write_bytes(b"video")
     seen: list[Path] = []

{classifyre_cli-0.4.58 → classifyre_cli-0.4.60}/uv.lock

@@ -830,7 +830,7 @@ wheels = [
 
 [[package]]
 name = "classifyre-cli"
-version = "0.4.58"
+version = "0.4.60"
 source = { editable = "." }
 dependencies = [
     { name = "beautifulsoup4" },
@@ -1347,7 +1347,7 @@ youtube = [
 
 [[package]]
 name = "classifyre-schemas"
-version = "0.4.58"
+version = "0.4.60"
 source = { editable = "../../packages/schemas" }
 dependencies = [
     { name = "fastjsonschema" },

classifyre_cli-0.4.58/.turbo/turbo-build.log

@@ -1,3 +0,0 @@
-$ uv sync
-Resolved 300 packages in 202ms
-Checked 51 packages in 3ms